24 TRANSPORTATION RF,SEARCH RF,CORD 1433

Effect of With-Flow Bus Lanes on Bus Travel Times

AMER 5. 5HALABY AND RICHARD M. SOBERMAN

Improvements in bus performance due to introduction of reserved bus in total travel time ranged from 0 to 15 percent, 5 to 25 percent, lanes have traditionally been evaluated in terms of savings in total and 7 percent, respectively (2). Discrepancies in the results of travel time. Little attention is usually paid to changes in individual these and other studies provide few guidelines for the expected segment times (i.e., travel times between consecutive bus stops along the bus route). An approach is presented that investigates the •effect change since, generally, the overall evaluations do not examine of an urban reserved bus lane on bus travel time on individual seg­ the impact on individual segments (i.e., sections between consec­ ments. Subsequently, the change in segment time is related to char­ utive bus stops). An extensive overview of bus priority experi­ acteristics and traffic regulations at respective segments. Data were ences in North American and European cities is presented else­ obtained by analysis of videotapes recorded before and after the in­ where (2,4). troduction of exclusive curb bus lanes on a major arterial road in In this study, the impact of with-flow bus lanes is investigated downtown Toronto. The data indicate that time savings are most likely to occur on segments where buses previously experienced consider­ by analyzing segments of a bus route individually. An attempt is able congestion, as well as at traffic signals, especially when bus stops made to relate changes in any one segment to traffic regulations are arranged with one on the near side and the next on the far side and characteristics of the particular segment. of their respective intersections. However, these time savings generate To study the effect of with-flow bus lanes on travel times, it is additional ridership, resulting in longer dwell times at stops and a essential to analyze conditions both before and after implemen­ corresponding overall increase in total travel time. Thus, the percep­ tation of the priority scheme. This can be accomplished by pre­ tion of transit service improvement may have more impacts on rider­ ship change than any substantive change in performance. The results paring time-space tables from which travel times for each segment of the study suggest opportunities for using reserved bus lanes on a can then be extracted. In obtaining the comparison results, statis­ more selective basis along a particular route and the need to reconsider tical tests should be used to determine whether the difference is whether taxis should be permitted to use these lanes. significant or whether the change could have occurred simply be­ cause of inherent variations. This is the approach used in com­ paring before and after observations for the Bay Street Urban Throughout North America, reduced dependence on public transit Clearway in Toronto. has led to increased road congestion with corresponding delays and costs. For example, the daily cost of delay in the United States in 1984, on the freeway system alone, is estimated to have ex­ ceeded $1.2 billion (1). By contrast, for any reasonable load fac­ BAY STREET URBAN CLEARWAY AND DATA tor, buses contribute relatively little to congestion. According to COLLECTION one study, a bus can carry 20 times as many passengers as a car and contributes only 3 times as much to congestion (2). As a Street and Service Characteristics result, bus priority schemes have attracted attention as a means of reducing bus delays due to traffic congestion in order to enhance Bay Street, one of the central corridors in downtown Toronto, the attractiveness of transit. extends north-south from Davenport Road to Queens Quay, as Although there are a variety of bus priority schemes, this paper shown in Figure 1. There are two lanes in each direction. Transi is concerned with urban streets along which curb lanes are devoted service consists of two overlapping bus routes, a main route sup­ to bus use, referred to as "with-flow bus lanes." The implemen­ plemented by a "short tum" route during morning and evenin tation of such preferential treatment is generally believed to result peak periods. In the morning period, most riders board buses a in an improvement in total bus travel time, taken as the best single Bloor Street, transferring from the subway to access employmen indicator of level of service. Previous applications of with-flow activities to the south. In the afternoon, the direction of flo bus lanes have shown a wide range of changes in total travel time. reverses. For example, a dual-width exclusive bus lane was introduced Most of the 17 bus stops covered by the service are located o on Madison Avenue in midtown Manhattan and before and after near sides of signalized intersections, and a few are located on fa observations were obtained for the entire length of this facility sides (Figure 2). Throughout this paper, bus stops are by defaul (3). During the p.m. peak hour, average total travel time decreased located on near side (at traffic signals), unless otherwise stated. by 45 percent, from approximately 18 min to less than 10 min. Canadian examples include bus lanes on Albert/Slater and Rideau streets in Ottawa and Eglinton Avenue in Toronto, where changes Project Implementation

Department of Civil Engineering, University of Toronto, 35 St. George On October 29, 1990, the city of Toronto initiated the dedicatio Street, Toronto, Ontario MSS 1A4 Canada. of curb lanes to buses, taxicabs, right-turning vehicles, and bicy Bloor Street

Front Street miles

LA K E

Sec A

FIGURE 1 Bay Street in metropolitan Toronto. 26 TRANSPOR'IATION RESEARCH RECORD 1433

Data Collection

Data were collected by continuous on-board.video camera filming through the bus windshield. The camera is equipped with a stop­ watch that indicates elapsed time, thereby allowing the travel time by segment to be determined. Before implementation, the operation of traffic control devices (i.e., tum prohibitions and parking) .varied throughout the day. Traffic flows and ridership also vary by time of day, as well as by directfon. For these reasons, filming was carried out both be­ fore and after implementation at three fixed times, namely, 8:00 a.m., 2:00 p.m., and 4:30 p.m., to represent the morning peak, off­ peak, and evening peak periods, respectively. For each period, one southbound and one northbound trip were filmed per day. Filming was carried out on Tuesdays, Wednesdays, and Thursdays to avoid irregularities usually associated with weekends, Mondays, and Fri­ days. The sample sizes for the before and after periods are pre­ sented in Table 1.

Data Preparation

T bus stop From the data collected, each of the southbound and northbound trips during the three periods (i.e., morning, off-peak, and eve­ • traffic signal ning), before and after project implementation, was separated into FIGURE 2 Illustration of segments with different bus stop . individual dwell times (at each stop), individual travel times from locations. each stop and/or traffic signal to the following stop and/or traffic signal, and individual signal times (i.e., delay time at each traffic signal). Finally, the data were entered for computer analysis. des. The project is documented by the Department of Public Works (5) and summarized as follows: ANALYSIS OF AGGREGATE TIMES 1. Almost 3 km of the southbound and northbound curb lanes on the Bay Street are reserved for public transit motor vehicles, Before turning to the detailed analysis of segment times, this sec­ taxicabs, right-turning vehicles, and bicycles only from 7:00 a.m. tion examines the change in bus performance on the basis of total to 7:00 p.m. except on Saturdays, Sundays, and public holidays. travel time. Total travel time for a southbound trip is defined as Seventeen bus stops in each direction are served by the reserved the time from the moment that doors open at the Bloor bus stop lanes. There is no break during the off-peak period because bus to allow for passenger boarding and alighting until the moment delays during this period are of the same order as those during that doors are closed at the Union Station bus stop in the south. the afternoon peak period (5). Most origins and destinations of passenger trips lie along this 2. Stopping, except for transit vehicles, is prohibited from 7:00 section. Total travel time for a northbound trip is defined similarly. a.m. to 7:00 p.m. except on Saturdays, Sundays, and public holi­ Total travel time includes total running and total dwell times. days. Parking is prohibited at all times on both sides of Bay Street. The change in total running time is considered a better measure Stopping was permitted before implementing this project at all of change in overall performance than total travel time because times, but parking was permitted during off-peak periods only. total travel time may increase because of increases in dwell time 3. Some new tum prohibitions were introduced. attributable to increased ridership (which itself, of course, is a 4. The reserved lanes are identified by overhead signs and pave­ positive result). ment markings of white painted diamonds with the message 7 To study the change in any of these three time measures after A.M.-7 P.M., MON-FRI, NO CARS-TRUCKS. project implementation, the t-test on two population means is

TABLE 1 Sizes of Before and After Samples

Number of Trips

Morning Mid-day Evening

Period SB0 NB SB NB SB NB 'Before' 16 13 8 7 14 14 'After' 12 12 IO 9 7 IO

0 SB = Southbound; NB = Northbound. Shalaby and Soberman 27

used. The t-test was carried out at a 5 percent level of significance findings of another study that reported an overall increase in rider­ as recommended for traffic studies (6). The results of all tests are ship by 25 percent (7). During the midday period, ridership in the given in Table 2 and discussed here by direction of travel. The southbound direction increased by 45.7 percent, from 45.8 to 66.7 mark (j) in the tables indicates that the null hypothesis tested is on-passengers per trip, as shown in Figure 3. rejected at the 5 percent level of significance, implying that the random variable has either decreased or increased. Northbound Direction

Since traffic in the northbound direction had not been affected by Southbound Direction construction, the results shown in Table 2 for this case provide a more reliable measure of the impact of the reserved lane on the Table 2 indicates that the means of total travel time and total overall bus performance. As shown, during the morning and mid­ running time decreased significantly during the three periods stud­ day periods, none of the means of the three variables changed ied. However, because of road construction activity that affected significantly after project implementation. During the morning pe­ travel times at southern segments before project implementation, riod, traffic is very light in the northbound direction and parking the changes in total travel and total running times for the south­ along Bay Street was already prohibited before the bus lane was bound direction do not represent the effect of the reserved lane introduced. Thus, the results pertaining to this period agree with alone on bus performance. In other words, without this construc­ the a priori expectation of changes in the three variables. tion activity, travel times before project implementation would Although parking was permitted during the midday period be­ undoubtedly have been lower than those recorded. fore project implementation, the expected improvement in bus As indicated in Table 2, the mean of total dwell time increased performance after introducing the exclusive lane, accompanied by by 44.8 percent during the midday period, with no significant parking prohibition, did not occur. Total dwell time did not change changes occurring during other periods. The increase in total dwell significantly, yet ridership shows an increase comparable to the time is attributable to increased ridership. To investigate the in­ case of midday, southbound period, as shown in Figure 3. crease in ridership, the numbers of passengers boarding and alight­ During the evening period, the mean of the total travel time ing at each stop were observed. Passengers board from the front increased significantly by 7.4 percent, while the mean of the total door and alight from either the front or rear doors. Since filming running time remained unchanged. However, the mean of the total was carried out from the front seat, passengers boarding and dwell time increased by 61.8 percent, which explains why the alighting from the front door were videotaped and subsequently mean of the total travel time increased. Corresponding increase in counted when the tapes were viewed. The number of passengers ridership is shown in Figure 3. alighting from the rear were counted manually at each stop and dictated into the camera microphone during taping. Figure 3 de­ 'picts the changes in ridership, measured by the total number ~f Conclusions Related to Aggregate Times on-passengers per bus trip along the bus lane for four of the six cases studied; ridership in the other two cases is minimal. The The results for the northbound direction during the evening period exhibit shows a general increase in ridership that agrees with the reveal the weaknesses of studying the change in bus performance

TABLE 2 Results of t-Tests on Aggregate Time Means

Morning Southbound Northbound

x,a X2 Change x, X2 Change (%) (%) Total Travel Time 18.2 17 ,/ -6.7 15.1 14.8 Total Running Time 13.9 12.2 ,/ -12.2 12.3 11.9 Total Dwell Time 4.3 4.8 2.8 2.9 Mid-day Total Travel Time 21.5 18.2 ,/ -15.3 17.4 173 Total Running Time 18.2 13.4 ,/ -26.2 14.2 12.6 Total Dwell Time 3.3 4.8 ,/ +44.8 3.3 4.6 Evening Total Travel Time 20.5 18.2 ,/ -11.6 19.3 20.7 ,/ +7.4 Total Running Time 17.1 14.l ,/ -17.8 15.5 14.6 Total Dwell Time 3.4 4.1 3.7 6.1 ,/ +61.8

0 x1 and x2 are sample averages (in minutes) 'before' and 'after', respectively. 28 TRANSPORTATION RESEARCH RECORD 1433

110

100 ~ before - after

90

;n 80 ~ ~G ~ (fl (fl 70 ~ 0.. I i: 0 60 0 Gi 50 ~ 3 z 40 ~ ~ 30

20

10

0 Morning, SB Mid-day, SB Mid-day, l\'B Evening, NB

FIGURE 3 Average total number of on-passengers per trip along the bus lane.

on the basis of total travel time change. According to the analysis, carded from the analysis, except for the morning period when bus performance during the evening period deteriorated signifi­ southbound traffic is relatively light at that particular section of cantly, after introducing the bus lane, whereas the mean of the Bay Street. total running time, a more precise measure of bus performance, Bus time mean, in the southbound direction during the morning did not change. period, decreased significantly after project implementation at For the northbound direction, results indicate that total running only 5 of the 15 segments studied, as indicated in Table 3. Parking time did not change after project implementation during any of and turning prohibitions at these five segments were already in the periods studied. The reasons for this result are unclear since, force before lane introduction. Inspection of these segment times thus far, the impacts of the reserved bus lane on individual seg­ shows that most savings occurred at four traffic signals. Examples ment running times have been ignored. Clearly, measuring total include signal times, which decreased from 21.7 to 3.2 sec and travel time alone does not help explain differential changes in the from 12.1 to 0.4 sec. two basic components (total running and total dwell times). More­ over, total running time does not account for different segment characteristics along the entire route.

TABLE 3 Results of t-Tests on Segment Time Means

CHANGES IN SEGMENT TIMES Sample Average (sec)

'Before' 'After' Change(%) Segment time is the time taken to travel between two successive bus stops, excluding dwell time. For a single segment, it is the 48.6 38.2 -21.4 elapsed time from when the doors are closed at the upstream stop 26.7 23.1 -13.5 until they are opened at the next stop. Signal time (i.e., bus delay Southbound, Morning 47.7 28.1 -41.1 at a traffic signal) is included if encountered during this period. t-tests are carried out for all segment times of southbound and 23 19.4 -15.6 northbound trips during the three periods studied. The results for 70 61.7 -11.8 segment time means that changed at the 5 percent significance level are presented in Table 3. The detailed analysis and results 86.5 51.5 -40.5 Southbound, Mid-day are presented more fully els~where on a segment-by-segment ba­ 44.7 26.6 -40.5 sis, for each direction, and by the three basic periods (A. S. Shal­ 45.2 58.5 +29.4 aby, unpublished data). Only a few of the more general observa­ tions are summarized herein. 66.l 85.6 +29.5 Northbound, Evening Construction, as noted previously, was taking place at a south­ 68.2 31.4 -53.9 ern intersection on the Bay Street before lane introduction. As a 121.1 77.4 -36.1 result, the four southbound segments that were affected are dis- Shalaby and Soberman 29

At one signalized intersection, where the bus stop is on the near located, and alight at the northern segments, especially at the side, before lane introduction, buses would usually encounter large Bloor stop, which is a transfer point between bus and subway queues that would prevent buses from boarding passengers during services. The results given in Table 3 indicate that bus time mean the red signal. As a result, buses would generally join the queue increased at 1 and decreased at 2 of the 16 segments studied. The until it dissipated, board and alight passengers during the green increase in segment time mean is due to considerable increase in signal, and were then forced to wait during the following red time. time mean (44.2 to 61.3 sec) for the traveled distance of this After lane introduction, queues are much shorter than before and segment (i.e., the segment time excluding the signal time at the formed by light-volume, right:.tuming vehicles Thus, buses can upstream intersection). In fact, thorough inspection showed that make use of the red signal to board and alight passengers; con­ the number of right-turning vehicles at the downstream intersec­ sequently, the need to wait for more than one red signal becomes tion increased considerably, leading to longer queues that delayed less likely. buses after lane introduction. The increased number of right­ The phenomenon of time savings at traffic signals is more pro­ tuming vehicles is possibly due to the prohibition on right turns nounced when the location of bus stops is such that one on the at the Bloor intersection and, consequently, the shift of right­ near side with the following stop on the far side of their respective tuming movements to other intersections. Furthermore, according intersections, an arrangement known as Von Stein's law of transit to one study, the number of taxis increased by 8 percent on the stop locations (8). The bus stops at four consecutive intersections Bay Street to take advantage of the reserved lane (7). Thus, taxis to the north of Bay Street constitute a series of alternative stops. attracted from alternatives to the Bay Street cause high delays at The results show that time means at two of the three segments intersections where they leave for their destinations. decreased significantly, mainly because of shorter delay times at Introduction of the bus lane, together with the prohibition of traffic signals, as noted earlier. This phenomenon also occurred at right turns at one intersection, relieved to a large extent the con­ the second series of alternative stops at three intersections south siderable delays experienced by buses in the two northern seg­ of the Bay Street. It is more pronounced when right-turning traffic ments where significant decrease in time mean occurred. The bus is either prohibited or very light. stops in the two consecutive segments are alternate, which also During the midday period, travel time mean decreased at one contributed to time savings, as explained earlier. segment and increased at another, as shown in Table 3. Inspection Effect on automobile times in non-bus-lanes as well as sur­ of the first segment indicates that the significant decrease in seg­ rounding streets was reported to be insignificant, except in six ment time mean is due to parking prohibition after lane introduc­ sections, one in Bay Street and five in neighboring streets, where tion. Although traffic congestion due to parking was also pro­ travel time decreased significantly (7). nounced at another segment during the before period, change in the segment time mean was obscured by illegal parking and stop­ ping because of ineffective police enforcement after lane intro­ SUMMARY AND RECOMMENDATIONS duction at that segment. When delays due to illegal parking and stopping (by cars, taxis, etc.) were eliminated from observations Change in bus performance, following the introduction of with­ (e.g., waiting time for a bus behind a vehicle stopping or parking flow bus lanes on urban streets, has usually been evaluated on the was excluded from observations), segment time mean decreased basis of total travel time change. This paper shows that total travel significantly by 40.5 percent. time is not the best measure of change in bus performance because The significant increase in time mean in the other segment, its components (i.e., running times and dwell times) may vary during the midday period in the southbound direction, is attributed considerably. Total running time also attributes the change in per­ to the considerable increase in signal time mean at the upstream formance to the overall characteristics of the street, with no focus intersection (2.2 to 17 sec), which may be a result of the increase on individual segments having different characteristics and traffic in dwell time at the near-side bus stop, causing buses to wait more regulations. As a result, reasons for changes, if any, are not fully often during the red signal after using the green time for boarding. explained. Automobile volume, measured for both directions combined at The analysis carried out in this study leads to the following that segment, was reported to have increased after project imple­ conclusions: mentation by 13 percent, from 1,310 to 1,480 (7). In three of the six cases studied-namely, southbound-evening, • The bus lane has little impact on bus performance during off­ northbound-morning, and northbound-midday-buses at the 16 peak periods and when traffic is light. segments of the route experienced no significant change in travel • Prohibition of parking, only at previously congested seg­ time mean. In one case, although parking prohibitions were ap­ ments, improves bus performance at those segments. plied after lane introduction, there was no effect on bus perfor­ •Time savings occur at traffic signals (especially at segments mance, largely because of relatively light traffic. In the other two accommodating alternative stops) and on previously congested cases, parking and stopping regulations were similar before and segments. after, while some new tum prohibitions were introduced after lane •Right-tum prohibitions improve bus performance consider­ introduction. However, no improvement occurred because of light ably. However, caution should be paid to the adverse impact of traffic. It is concluded, therefore, as might be expected, that dedi­ diversions of traffic to alternative intersections at which right turns cation of the curb lane for transit use has no effect on segment are permitted imes when traffic volume is light, even though turning move­ • Police enforcement is an important factor in achieving im­ ents, stopping, and parking prohibitions favor bus performance. provements to bus performance, particularly on congested As noted earlier, traffic flow in the northbound direction rep- segments. esents the peak flow during the evening period. Most riders board • The use of reserved lanes by taxis diverted from other streets uses at the southern stops, where major employment centers are contribute to bus delays. 30 TRANSPORTATION RESEARCH RECORD 1433

• Ridership generally increases after introducing the lane, even REFERENCES without improvements in travel time. 1. Proc., 2nd National Conference on High-Occupancy Vehicles. Houston, The last finding is noteworthy, since it appears that ridership in­ Tex., Oct. 1987. 2. NATO Committee on Challenges of Modern Society. Bus Priority Sys­ creased because of the perception of an enhanced service by es­ tems. Transport and Road Research Laboratory, CCMS, Vol. 45, 1976. tablishing an exclusive lane, even though total travel time, in one 3. Schwartz, S. I., A. Hollander, C. Louie, and R. Amoruso. Madison case, increased. In a user attitudinal survey,· for example, 91 per­ Avenue Dual-Width Bus Lane Project. In Transportation Research Rec­ cent of the respondents expressed positive views of the project ord 854, TRB, National Research Council, Washington, D.C., 1982, and 85 percent claimed they have a reduced transit travel pp. 70-77. 4. Levinson, S. H., W. F. Hoey, D. B. Sanders, and F. H. Wynn. NCHRP time (7). Report 143: Bus Use of Highways: State of the Art. HRB, National For future projects the following should be taken into Research Council, Washington, D.C., 1973. consideration: 5. Bay Street Urban Clearway. Report c39csc90366:117. Department of Public Works, Toronto, Ontario, Canada, April 1990. 6. Bus Lanes and Busway Systems. Road Research Group, Organization • Dedicating curb lanes to bus use during peak periods only; for Economic Cooperation and Development, 1976. • Dedicating curb lanes to bus use on a selective basis, at con­ 7. The Bay Street Urban Clearway: Final Evaluation Report. Report gested segments only, preventing right turns where possible, and c39csc92149:117. Department of Public Works and the Environment, revising stop locations, to be alternative at those segments. Po­ Toronto, Ontario, Canada, Feb. 1992. 8. Vuchic, V. R. Urban Public Transportation Systems and Technology. lice enforcement, however, should be strict at those particular Prentice-Hall, Englewood Cliffs, N.J., 1981. segments; • Allowing taxis to use the ''jumping'' lanes (i.e., curb lanes at the segments at which buses are favored) should be considered more carefully because of potential adverse impacts on bus per­ formance; and • Allowing parking on lightly congested segments. Publication of this paper sponsored by Committee on Bus Transit Systems.